AVMD-Kriging砂轮寿命周期磨削性能在线监测方法

Online Monitoring Method for Grinding Performance of Grinding Wheel Life Cycle Based on A VMD-Kriging

为了对砂轮寿命周期磨削性能进行特征提取与智能识别,课题组提出了一种改进的变分模态分解算法与Kriging模型相结合的砂轮寿命周期磨削性能识别方法AVMD-Kriging。首先,通过人工鱼群算法和包络熵适应度函数来优化VMD,以解决VMD中本征模态函数分解个数k和惩罚因子α难以自适应确定的问题;再利用皮尔逊相关系数选取与原始信号相关性最高的本征模态函数并计算其样本熵值组成特征向量,将其输入Kriging模型进行砂轮寿命周期磨削性能识别;最后利用实验采集的声发射数据,将提出的AVMD-Kriging方法与传统的KNN模型、Tree模型进行对比。结果表明:AVMD-Kriging方法的识别准确率优于KNN模型和Tree模型,能有效提高砂轮寿命周期磨削性能的识别准确率,同时具有较好的泛化能力和鲁棒性。

For the problem of feature extraction and intelligent recognition of grinding wheel life cycle grinding performance,an improved variational modal decomposition method VMD combined with Kriging model was proposed to identify the grinding wheel life cycle grinding performance AVMD-Kriging.First,the VMD was optimized by artificial fish swarm algorithm and envelope entropy adaptation function to solve the problem of difficulty in adaptively determining the number and penalty factors of eigenmode function decompositions in VMD.Then,the Pearson correlation coefficient was used to select the eigenmodal function with the highest correlation with the original signal and calculate its sample entropy value to form a feature vector,which was input to the Kriging model for grinding wheel life cycle grinding performance identification.Finally,using the experimentally collected acoustic emission data,the proposed AVMD-Kriging method was compared with the traditional K-nearest neighbor algorithm model and decision Tree model.The results show that the recognition accuracy of the AVMD-Kriging method is superior to K-nearest neighbor algorithm model and the decision Tree model.AVMD-Kriging can effectively improve the recognition accuracy of grinding wheel life cycle grinding performance,and has good generalization ability and robustness.